One workspace for your team and every AI.
People, models, and tools — all in the same room, sharing one context.
A room where context lives.
Teammates, models, and tools all sit in the same Space — not in separate tabs, not behind copy‑paste, not lost in DMs. Everything anyone learns becomes the floor everyone else stands on next.
Your teammates, in one room. Conversation stays human‑default — AI replies only when @mentioned.
Claude, GPT‑5, Gemini, and more — invited as participants. Each one has full visibility into the thread.
GitHub, Linear, Slack, Notion, Gmail. Connect once; everyone in the Space can call them by @mention.
Type @claude for one answer. @gpt‑5 for another. Same thread.
Mix frontier models in one conversation. Image, video, and voice models share the same credit pool — no separate accounts, no separate billing, no lock‑in.
Claude
Gemini
Mistral
GrokConnect a tool once. The whole team @mentions it.
MCP connections are workspace‑wide. One person installs GitHub; the rest of the Space can ship PRs, file Linear issues, or pull a Notion doc into context without setup.
And Switchy's team memory is itself an MCP server — so the same context shows up automatically inside everyone's Claude Code, Cursor, Claude Desktop, and ChatGPT.
One URL per Space. Paste it into your editor's MCP config; what your team learned this morning is in your prompt this afternoon.
Start alone. Add teammates with one email.
Sign up and you have a full Space — every model, every MCP, your own private memory. Keep it that way for as long as you want. Nothing about Switchy assumes a team.
Type a teammate's email. They land in the same Space — same models, same tools, same memory. No migration, no export, no second tool to learn. Personal stays personal; the team only sees what you scope to them.
Build it into your own stack.
Every model, every MCP, and your team's memory layer is callable from your code. Use Switchy as a workspace — or as a backend.
One endpoint, every provider. Switch model mid‑run. Read or write to a Space's memory the same way a teammate would. Bearer auth, JSON in, JSON out.
The same memory you see in the app shows up in your API calls — and in every teammate's Claude Code, Cursor, and ChatGPT.
Read the API docs# pip install switchy
from switchy import Switchy
sx = Switchy(api_key="sx_…", space="team-product")
reply = sx.chat(
model="claude-sonnet-4.5",
message="What did we decide about onboarding step 3?",
tools=["github", "linear"],
)
print(reply.text)
# → Three-frame walkthrough; MCP step is now a soft prompt
# after the first thread (decided 2026-04-29 by @maya).